Premium
SU‐F‐303‐08: Four‐Dimensional MRI Method Based On Internal Respiratory Surrogate Derived by Dimensionality Reduction
Author(s) -
Uh J,
Hua C
Publication year - 2015
Publication title -
medical physics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.473
H-Index - 180
eISSN - 2473-4209
pISSN - 0094-2405
DOI - 10.1118/1.4925235
Subject(s) - preprocessor , diaphragm (acoustics) , artificial intelligence , computer science , dimensionality reduction , pattern recognition (psychology) , position (finance) , reduction (mathematics) , computer vision , mathematics , nuclear medicine , medicine , physics , finance , acoustics , loudspeaker , economics , geometry
Purpose: To investigate the feasibility of constructing 4D MRI using non‐navigator, image‐based internal respiratory surrogate derived by dimensionality reduction (DR). Methods: DR is a mathematical process that derives a series of scalar indices from the corresponding image frames. We proposed a novel image acquisition scheme (A) designed for the DR‐ derived internal surrogates (Wachinger, 2012) to better match the temporal pattern of previously established surrogates such as diaphragm position. The idea is to acquire a pair of slices alternatingly so the phase of a slice can be more accurately determined with additional information from the other slice. This scheme was compared with the sequential acquisition scheme (B), the navigator‐based approach (C), and a previously proposed method using body area as internal respiratory surrogate (Cai, 2011) (D). Correlation between internal surrogates from all schemes and the diaphragm position was studied using a publicly available 4D MRI dataset (von Siebenthal, 2007). Results: DR did not require any manual preprocessing. It took only a few minutes to process the entire 4D MRI data of 100 respiratory cycles with a common desktop computer. The respiratory surrogate signals were highly correlated with diaphragm positions. Correlation coefficients were 0.996, 0.993, 0.997, and 0.982 and mean absolute phase errors were 1.5%, 2.0%, 1.3%, and 3.3% for A, B, C and D, respectively. With the navigator‐based approach C, the sampling rate of the regions of interest was reduced by half. Among those without the need for navigators, the novel acquisition scheme A reflected diaphragm positions better than the sequential scheme B and body area scheme D. Conclusion: The proposed image‐based 4D MRI method with a novel acquisition scheme and the DR‐derived internal respiratory surrogate compared favorably to the navigator‐based approach and has twice the sampling rate. Although tested on sagittal slices, it is potentially applicable to any slice orientation.